grilling
Pass
Audited by Gen Agent Trust Hub on Nov 4, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied plans and instructs the agent to autonomously retrieve facts from the environment (filesystem, tools) using sub-agents, creating an attack surface for data exposure.
- Ingestion points: User-provided 'plan, decision, or idea' enters the context to initiate the questioning process.
- Boundary markers: No explicit delimiters or instructions are present to separate user input from the fact-finding logic for sub-agents.
- Capability inventory: The skill explicitly enables environmental access to the filesystem and tools via sub-agents.
- Sanitization: There is no specification for sanitizing user-provided ideas before they are used to determine which 'facts' the sub-agents should retrieve.
Audit Metadata